| name | fable-and-improve |
| description | Use when a Claude Code session should let Claude Fable plan and manage the goal, delegate implementation/review/improvement to the Codex design-and-implement workflow, then have Fable verify completion and loop on remaining TODOs. |
| metadata | {"short-description":"Fable plans, Codex implements, Fable verifies"} |
Fable And Improve
Use this skill when the user explicitly wants the fable-and-improve Riela
workflow or asks for a Fable-led plan with Codex implementation and review.
Package
- Package id:
fable-and-improve
- Workflow id:
fable-and-improve
- Parent backend:
claude-code-agent
- Parent model:
claude-fable-5
- Child workflow:
codex-design-and-implement-review-loop
- Required dependency package:
codex-design-and-implement-review-loop
This package intentionally provides only a Claude skill. It does not project a
Codex or Cursor usage skill.
Install
From a local riela-packages checkout:
riela package install fable-and-improve --source <riela-packages-checkout>/packages/fable-and-improve
User scope:
riela package install fable-and-improve --source <riela-packages-checkout>/packages/fable-and-improve --scope user
Validate
Project scope:
riela workflow validate fable-and-improve
riela workflow inspect fable-and-improve --output json
User scope:
riela workflow validate fable-and-improve --scope user
riela workflow inspect fable-and-improve --scope user --output json
Run
Use a concise workflowInput:
riela workflow run fable-and-improve \
--variables '{"workflowInput":{"requestedOutcome":"Implement the requested change, review it, and verify completion.","targetScope":"Describe the target files or feature area.","constraints":["Do not modify unrelated files."],"acceptanceCriteria":["The requested behavior works."],"verificationHint":"Run the smallest relevant verification."}}' \
--output jsonl
For a cheap connectivity check before using Fable, patch the parent Fable nodes
to a less expensive Claude Code model:
riela workflow run fable-and-improve \
--node-patch '{"fable-goal-plan":{"model":"claude-sonnet-4-5"},"codex-implementation-handoff":{"model":"claude-sonnet-4-5"},"codex-implementation-result":{"model":"claude-sonnet-4-5"},"fable-goal-review":{"model":"claude-sonnet-4-5"},"final-output":{"model":"claude-sonnet-4-5"}}' \
--variables '{"workflowInput":{"requestedOutcome":"Connectivity smoke test only.","targetScope":"No repository changes.","constraints":["Do not edit files."],"acceptanceCriteria":["The workflow starts and delegates correctly."],"verificationHint":"No-op smoke check."}}' \
--output jsonl
Then run the normal workflow with Fable for a small real task.
Output
Expect the final output to include:
status
goalAchieved
goalMarkdown
planMarkdown
delegatedWorkflowRuns
changedFiles
verificationEvidence
reviewEvidence
commitStatus
pushStatus
residualRisks
Maintainer Notes
After changing workflow files, prompts, mock scenarios, or this skill payload,
refresh digests:
bun .agents/skills/riela-package-release/scripts/update-package-digests.ts fable-and-improve